Given the current prevalence and impact of cervical cancer worldwide, many technological developments focused on automating the screening process have arisen recently. Nonetheless, there is still a clear need for affordable, portable and automated IoT-based solutions to expand the coverage of current cervical screening programs worldwide. This is particularly relevant for lower-resource countries, which account for 88% of all cervical cancer-related deaths. This work proposes a low-cost, smartphone-based microscopy device for the analysis of liquid-based cytology samples, through autonomous image acquisition and automated identification of cervical lesions. Different deep learning models for object detection were separately optimised and compared to select the most adequate network architecture. Transfer learning from a similar application domain - conventional cytology - was also investigated as a way of improving the robustness of the analysis pipeline, as well as overcoming the limitations of the mobile-acquired image dataset specifically collected and manually annotated by specialists under the scope of this work. In this process, a detection performance benchmark in the SIPAKMED dataset - test mean average precision (mAP) of 0.37798 and average recall (AR) of 0.63651 - was reported for the first time. Although further improvements are required for its integration in a computer-aided diagnosis system sufficiently reliable for deployment in a clinical context, the explored approach exhibits promising results (cross-validation mAP of 0.20315, AR of 0.46572 and analysis time of 4 minutes per cytological sample), corresponding to a step forward in the development of a cost-effective mobile IoT framework that supports cervical lesion screening.

Liquid-based cytology (LBC) plays a crucial role in the effective early detection of cervical cancer, contributing to substantially decreasing mortality rates. However, the visual examination of microscopic slides is a challenging, time-consuming, and ambiguous task. Shortages of specialized staff and equipment are increasing the interest in developing artificial intelligence (AI)-powered portable solutions to support screening programs. This paper presents a novel approach based on a RetinaNet model with a ResNet50 backbone to detect the nuclei of cervical lesions on mobile-acquired microscopic images of cytology samples, stratifying the lesions according to The Bethesda System (TBS) guidelines. This work was supported by a new dataset of images from LBC samples digitalized with a portable smartphone-based microscope, encompassing nucleus annotations of 31,698 normal squamous cells and 1395 lesions. Several experiments were conducted to optimize the model’s detection performance, namely hyperparameter tuning, transfer learning, detected class adjustments, and per-class score threshold optimization. The proposed nucleus-based methodology improved the best baseline reported in the literature for detecting cervical lesions on microscopic images exclusively acquired with mobile devices coupled to the µSmartScope prototype, with per-class average precision, recall, and F1 scores up to 17.6%, 22.9%, and 16.0%, respectively. Performance improvements were obtained by transferring knowledge from networks pre-trained on a smaller dataset closer to the target application domain, as well as including normal squamous nuclei as a class detected by the model. Per-class tuning of the score threshold also allowed us to obtain a model more suitable to support screening procedures, achieving F1 score improvements in most TBS classes. While further improvements are still required to use the proposed approach in a clinical context, this work reinforces the potential of using AI-powered mobile-based solutions to support cervical cancer screening. Such solutions can significantly impact screening programs worldwide, particularly in areas with limited access and restricted healthcare resources.

Rigorous and periodic monitoring of healing progression in chronic wounds is crucial for the proper management of this impactful health problem; however, manual assessment is highly subjective. Artificial Intelligence-based digital systems have arisen as a solution to automate the analysis of wound properties and reduce the variability of the assessment process at different levels. This study presents an automated approach for wound bed characterisation, using open wound detection and tissue segmentation algorithms to estimate the relative proportion of granulation, slough and eschar tissues. The impact of dataset composition on the performance of deep learning-based tissue segmentation models is investigated, along with the comparison of two model architectures (DeepLabV3+ with a ResNet50 backbone and UPerNet with a Swin Transformer backbone). Two private Wounds datasets are used and augmented with an external third-party external dataset. The results demonstrate that DeepLabV3+ outperforms UPerNet-Swin across all dataset combinations, establishing it as the preferred architecture in this case. Furthermore, incorporating all available datasets improves segmentation performance, underscoring the importance of data diversity. The best-performing model achieved Dice scores of $\mathbf{7 4. 6 5 \%}, \mathbf{6 2. 0 7 \%}$ and $\mathbf{6 9 \%}$ for granulation, slough, and eschar tissues, respectively, with corresponding mean absolute errors (MAE) of $\mathbf{1 6. 5 1 \%}, \mathbf{1 4. 6 8 \%}$, and $\mathbf{4. 2 6 \%}$ for tissue proportion estimation. A comparison of the standard deviation of the obtained MAE results with the ones reported for clinical experts in a similar task demonstrated that the proposed pipeline effectively decreased the variability of the estimated tissue percentages, providing a framework with the potential to streamline the wound monitoring process and increase its reproducibility.

Wound dressings and their proper management are crucial to the wound's recovery. This process can be time-consuming and requires special knowledge to be effective. In order to improve the monitorization and decision-making processes, this work proposes a framework based on state-of-the-art Deep Learning models for automating the acquisition and analysis of wound dressings. Its development was supported by a novel dataset of dressing images annotated by experts regarding dressing state and regions of interest. The two-step acquisition pipeline resorts to a RetinaNet model to detect the dressing region, along with a reference marker, further used by the image validation module to ensure that the images fulfil the clinical adequacy requirements, such as the presence of a minimum periwound area. On top of its robust detection performance, its mobile deployment demonstrated its ability to support and standardise the image acquisition task. The characterisation module analyses the dressing areas provided by the detection model, using a MobileNetV3Small model to classify if the dressing is usable or in need of change and achieving an F1 score of 0.778. Thus, this work provides a promising system to streamline the dressing monitoring process, constituting an important advancement in this field with the potential to empower caregivers and healthcare professionals.